[S] Survival analysis - frailty models and AIC

["Cirad" <moisan@telecomplus.sn>]

Hi everyone,

I'm working on survival models where a random effect is added (frailty
models).
I would like to extract the Akaike's information criterion (AIC) to
determine the best model.

Is the following function correct? Can I extract the number of
parameters 
estimated by the following code: 2*ncol(cox.obj$var)?

coxAIC <- function(cox.obj){
        -2 * (cox.obj$loglik[2] - ncol(cox.obj$var))
}


Example :

I've got three fixed effects and one random effect (village) :

nkolov.coxG <- coxph(Surv(duree, censure) ~ verm 
                                                + portee 
                                                + vacc 
                                                + 
frailty(village),singular.ok=T,
                                                 eps = 0.01, nkolov3)

coxAIC(nkolov.coxG)

The details of the AIC calculation is :

> nkolov.coxG$loglik[2]
[1] -1915.367

> ncol(nkolov.coxG$var)
[1] 3

> coxAIC(nkolov.coxG)
[1] 3836.734

Is there a better way to get the AIC value?
Thanks a lot for your help.

-- 
Arnaud Moisan
ISRA-LNERV / CIRAD-EMVT
BP 2057
Dakar Hann
Senegal

Tel:    (00 221) 832 49 02      (bureau)
Email : moisan@telecomplus.sn

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news